From 1c9e8db8e45b292b834d395bb000d94aa973fc99 Mon Sep 17 00:00:00 2001
From: Luvu182 <208665161+Luvu182@users.noreply.github.com>
Date: Wed, 4 Mar 2026 09:38:38 +0700
Subject: [PATCH] fix: OpenAI-compat SSE parsing, scanner error check,
 reasoning passback, and env var parity

- SSE: handle "data:" without space (Kimi and other providers)
- Add scanner.Err() check to detect stream read failures
- Echo reasoning_content for thinking models (Kimi, DeepSeek)
- Add Thinking field to Message struct for reasoning passback
- Add GOCLAW_OPENAI_BASE_URL env var override (parity with Anthropic)
---
 docker-compose.yml             |  1 +
 internal/config/config_load.go |  1 +
 internal/providers/openai.go   | 17 +++++++++++++++--
 internal/providers/types.go    |  1 +
 4 files changed, 18 insertions(+), 2 deletions(-)

diff --git a/docker-compose.yml b/docker-compose.yml
index 191cc388..7d4770ef 100644
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -35,6 +35,7 @@ services:
       - GOCLAW_ANTHROPIC_API_KEY=${GOCLAW_ANTHROPIC_API_KEY:-}
       - GOCLAW_ANTHROPIC_BASE_URL=${GOCLAW_ANTHROPIC_BASE_URL:-}
       - GOCLAW_OPENAI_API_KEY=${GOCLAW_OPENAI_API_KEY:-}
+      - GOCLAW_OPENAI_BASE_URL=${GOCLAW_OPENAI_BASE_URL:-}
       - GOCLAW_MINIMAX_API_KEY=${GOCLAW_MINIMAX_API_KEY:-}
       - GOCLAW_GROQ_API_KEY=${GOCLAW_GROQ_API_KEY:-}
       - GOCLAW_DEEPSEEK_API_KEY=${GOCLAW_DEEPSEEK_API_KEY:-}
diff --git a/internal/config/config_load.go b/internal/config/config_load.go
index 02729206..989b48b8 100644
--- a/internal/config/config_load.go
+++ b/internal/config/config_load.go
@@ -95,6 +95,7 @@ func (c *Config) applyEnvOverrides() {
 	envStr("GOCLAW_ANTHROPIC_API_KEY", &c.Providers.Anthropic.APIKey)
 	envStr("GOCLAW_ANTHROPIC_BASE_URL", &c.Providers.Anthropic.APIBase)
 	envStr("GOCLAW_OPENAI_API_KEY", &c.Providers.OpenAI.APIKey)
+	envStr("GOCLAW_OPENAI_BASE_URL", &c.Providers.OpenAI.APIBase)
 	envStr("GOCLAW_OPENROUTER_API_KEY", &c.Providers.OpenRouter.APIKey)
 	envStr("GOCLAW_GROQ_API_KEY", &c.Providers.Groq.APIKey)
 	envStr("GOCLAW_DEEPSEEK_API_KEY", &c.Providers.DeepSeek.APIKey)
diff --git a/internal/providers/openai.go b/internal/providers/openai.go
index 7b903808..18d5c174 100644
--- a/internal/providers/openai.go
+++ b/internal/providers/openai.go
@@ -105,10 +105,13 @@ func (p *OpenAIProvider) ChatStream(ctx context.Context, req ChatRequest, onChun
 	scanner := bufio.NewScanner(respBody)
 	for scanner.Scan() {
 		line := scanner.Text()
-		if !strings.HasPrefix(line, "data: ") {
+		if !strings.HasPrefix(line, "data:") {
 			continue
 		}
-		data := strings.TrimPrefix(line, "data: ")
+		// SSE spec allows both "data: value" and "data:value" (space is optional).
+		// Some providers (e.g. Kimi) omit the space after the colon.
+		data := strings.TrimPrefix(line, "data:")
+		data = strings.TrimPrefix(data, " ")
 		if data == "[DONE]" {
 			break
 		}
@@ -174,6 +177,11 @@ func (p *OpenAIProvider) ChatStream(ctx context.Context, req ChatRequest, onChun
 
 	}
 
+	// Check for scanner errors (timeout, connection reset, etc.)
+	if err := scanner.Err(); err != nil {
+		return nil, fmt.Errorf("%s: stream read error: %w", p.name, err)
+	}
+
 	// Parse accumulated tool call arguments
 	for i := 0; i < len(accumulators); i++ {
 		acc := accumulators[i]
@@ -217,6 +225,11 @@ func (p *OpenAIProvider) buildRequestBody(model string, req ChatRequest, stream
 			"role": m.Role,
 		}
 
+		// Echo reasoning_content for assistant messages (required by Kimi, DeepSeek when thinking is enabled)
+		if m.Thinking != "" && m.Role == "assistant" {
+			msg["reasoning_content"] = m.Thinking
+		}
+
 		// Include content; omit empty content for assistant messages with tool_calls
 		// (Gemini rejects empty content → "must include at least one parts field").
 		if m.Role == "user" && len(m.Images) > 0 {
diff --git a/internal/providers/types.go b/internal/providers/types.go
index 929b9519..4450a601 100644
--- a/internal/providers/types.go
+++ b/internal/providers/types.go
@@ -76,6 +76,7 @@ type ImageContent struct {
 type Message struct {
 	Role       string         `json:"role"`                  // "system", "user", "assistant", "tool"
 	Content    string         `json:"content"`
+	Thinking   string         `json:"thinking,omitempty"`    // reasoning_content for thinking models (Kimi, DeepSeek, etc.)
 	Images     []ImageContent `json:"images,omitempty"`      // vision: base64 images
 	ToolCalls  []ToolCall     `json:"tool_calls,omitempty"`
 	ToolCallID string         `json:"tool_call_id,omitempty"` // for role="tool" responses